Balak, Patryk; Ł, Agnieszka ; ń, Kamil Choroma; Luckner, Marcin Influence of Augmentation of UAV Collected Data on Deep Learning Based Facade Segmentation Task Inproceedings ć, Lukovi I; ć, Bjeladinovi S; ć, Delibaši B; ć, Bara D; Iivari, N; Insfran, E; Lang, M; Linger, H; Schneide, C (Ed.): Empowering the Interdisciplinary Role of ISD in Addressing Contemporary Issues in Digital Transformation: How Data Science and Generative AI Contributes to ISD (ISD2025 Proceedings), pp. 1–5, University of Gdańsk, Department of Business Informatics & University of Belgrade, Faculty of Organizational Sciences., Belgrade, 2025. Abstract | Links | BibTeX | Tagi: augmentation, Computer vision, segmentation, unmanned aerial vehicles @inproceedings{Balak2025,
title = {Influence of Augmentation of UAV Collected Data on Deep Learning Based Facade Segmentation Task},
author = {Patryk Balak and Agnieszka {\L}ysak and Kamil Choroma{\'{n}}ski and Marcin Luckner},
editor = {I Lukovi{\'{c}} and S Bjeladinovi{\'{c}} and B Deliba\v{s}i{\'{c}} and D Bara{\'{c}} and N Iivari and E Insfran and M Lang and H Linger and C Schneide},
url = {https://aisel.aisnet.org/isd2014/proceedings2025/datascience/34/},
doi = {10.62036/ISD.2025.64},
year = {2025},
date = {2025-01-01},
booktitle = {Empowering the Interdisciplinary Role of ISD in Addressing Contemporary Issues in Digital Transformation: How Data Science and Generative AI Contributes to ISD (ISD2025 Proceedings)},
pages = {1--5},
publisher = {University of Gda\'{n}sk, Department of Business Informatics & University of Belgrade, Faculty of Organizational Sciences.},
address = {Belgrade},
abstract = {Data augmentation is crucial for image segmentation, especially in transfer learning with limited data, however it can be costly. This study examines the cost-benefit of augmentation in facade segmentation using unmanned aerial vehicles (UAV) data. We analysed how dataset size and offline augmentation impact classification quality and computation using DeepLabV3+ architecture. Expanding the dataset from 5 to 480 thousand tiles improved segmentation efficiency by an average of 3.7%. Beyond a certain point, further dataset expansion yields minimal gains, in our case, just 1%, on average, after segmentation accuracy plateaued at around 76%. These findings help avoid the computational and time costs of ineffective data augmentation.},
keywords = {augmentation, Computer vision, segmentation, unmanned aerial vehicles},
pubstate = {published},
tppubtype = {inproceedings}
}
Data augmentation is crucial for image segmentation, especially in transfer learning with limited data, however it can be costly. This study examines the cost-benefit of augmentation in facade segmentation using unmanned aerial vehicles (UAV) data. We analysed how dataset size and offline augmentation impact classification quality and computation using DeepLabV3+ architecture. Expanding the dataset from 5 to 480 thousand tiles improved segmentation efficiency by an average of 3.7%. Beyond a certain point, further dataset expansion yields minimal gains, in our case, just 1%, on average, after segmentation accuracy plateaued at around 76%. These findings help avoid the computational and time costs of ineffective data augmentation. |